How this calculation works
Estimate the plausible range of one new observation from a normal population.
Prediction bounds=x̄±t* s√(1+1/n)A worked example
With mean 10, SD 2, n=25 and 95% confidence, the bounds are about 5.79 and 14.21.
Good to know
The new observation must be independent and from the same population.
Reference: OpenStax Introductory Statistics (opens in a new tab) ↗
Work through the 95% prediction interval
Suppose 25 independent observations have a mean of 10 and a sample standard deviation of 2. A two-sided 95% interval uses the 97.5th percentile of Student’s t distribution with 24 degrees of freedom, about 2.064.
The prediction standard error is 2 × √(1 + 1/25), about 2.040. Multiplying by the critical value gives a margin of about 4.21. Subtract and add this margin to the mean to get bounds of 5.79 and 14.21, in the same units as the original observations.
Choose the right interval for your question
A confidence interval for the population mean uses s/√n. A prediction interval for one future observation also includes the variability of the observation itself, which is why its standard error uses √(1 + 1/n). More data make the estimated mean more precise, while future individual observations still vary.
The Student t calculation assumes independent observations from the same normal population. If you are predicting tomorrow’s reading in a trending time series, the sample mean and standard deviation alone do not account for the trend or dependence.
A few questions, answered
Should I round the inputs?
Keep the full precision of your sample mean and sample standard deviation if available, and round the two bounds for display.
Does this interval cover several future observations at once?
It is designed for one future observation. Simultaneous coverage for several observations requires a different calculation.